Predicting Future Hydrocarbon Production using Neural Networks on Near Wellbore Seismic Volumes

Oil companies make money leveraging their land position and producing oil and gas. Today, with Tier 1 acreage becoming harder and harder to come by, and any project requiring hydraulic fracturing coming under scrutiny. Acquiring the best leases, putting as few wells as possible in the best and safest locations and drilling them to optimal lengths to maximize production and minimize surface disturbances becomes key. As an industry, we have a lot of data these days, and especially so in more difficult areas. Data science is a tool that allows us to efficiently distil all this data into useful information. Since hydrocarbon production is what we are ultimately after, we have developed a method to predict the production of wells, including horizontal wells, away from existing well control, when the predictors are available in the vicinity of the wells yet to be drilled. This paper shows a case study where we use existing production to predict production elsewhere by carefully selecting attributes to put into a Neural Network (NN). The results show that there are significant areas of Tier 1, Tier 2 and Tier 3 acreage. Once you get away from the best of these, and even within them, there are many opportunities to optimize production.